pauloricardomg commented on code in PR #4558: URL: https://github.com/apache/cassandra/pull/4558#discussion_r2738435849
########## src/java/org/apache/cassandra/repair/autorepair/AutoRepairUtils.java: ########## @@ -0,0 +1,1419 @@ +/* + * Licensed to the Apache Software Foundation (ASF) under one + * or more contributor license agreements. See the NOTICE file + * distributed with this work for additional information + * regarding copyright ownership. The ASF licenses this file + * to you under the Apache License, Version 2.0 (the + * "License"); you may not use this file except in compliance + * with the License. You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ +package org.apache.cassandra.repair.autorepair; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.Arrays; +import java.util.Collection; +import java.util.Collections; +import java.util.Comparator; +import java.util.HashMap; +import java.util.HashSet; +import java.util.Iterator; +import java.util.List; +import java.util.Map; +import java.util.Optional; +import java.util.Set; +import java.util.TreeSet; +import java.util.UUID; +import java.util.concurrent.TimeUnit; +import java.util.concurrent.atomic.AtomicBoolean; +import java.util.stream.Collectors; +import java.util.stream.Stream; + +import javax.annotation.Nullable; + +import com.google.common.annotations.VisibleForTesting; +import com.google.common.base.MoreObjects; +import com.google.common.collect.Lists; +import com.google.common.collect.Sets; + +import com.clearspring.analytics.stream.cardinality.CardinalityMergeException; +import com.clearspring.analytics.stream.cardinality.HyperLogLogPlus; +import com.clearspring.analytics.stream.cardinality.ICardinality; +import org.apache.cassandra.dht.Range; +import org.apache.cassandra.dht.Splitter; +import org.apache.cassandra.dht.Token; +import org.apache.cassandra.io.sstable.format.SSTableReader; +import org.apache.cassandra.io.sstable.metadata.CompactionMetadata; +import org.apache.cassandra.io.sstable.metadata.MetadataType; +import org.apache.cassandra.locator.EndpointsByRange; +import org.apache.cassandra.locator.EndpointsForRange; +import org.apache.cassandra.locator.LocalStrategy; + +import org.apache.cassandra.utils.CassandraVersion; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import org.apache.cassandra.config.DatabaseDescriptor; +import org.apache.cassandra.cql3.QueryOptions; +import org.apache.cassandra.cql3.QueryProcessor; +import org.apache.cassandra.cql3.UntypedResultSet; +import org.apache.cassandra.cql3.statements.ModificationStatement; +import org.apache.cassandra.cql3.statements.SelectStatement; +import org.apache.cassandra.db.ConsistencyLevel; +import org.apache.cassandra.db.Keyspace; +import org.apache.cassandra.db.marshal.UTF8Type; +import org.apache.cassandra.db.marshal.UUIDType; +import org.apache.cassandra.gms.Gossiper; +import org.apache.cassandra.locator.AbstractReplicationStrategy; +import org.apache.cassandra.locator.InetAddressAndPort; +import org.apache.cassandra.locator.NetworkTopologyStrategy; +import org.apache.cassandra.locator.RangesAtEndpoint; +import org.apache.cassandra.locator.Replica; +import org.apache.cassandra.metrics.AutoRepairMetricsManager; +import org.apache.cassandra.schema.Schema; +import org.apache.cassandra.schema.SchemaConstants; +import org.apache.cassandra.schema.SystemDistributedKeyspace; +import org.apache.cassandra.schema.TableMetadata; +import org.apache.cassandra.schema.ViewMetadata; +import org.apache.cassandra.serializers.SetSerializer; +import org.apache.cassandra.serializers.UUIDSerializer; +import org.apache.cassandra.service.AutoRepairService; +import org.apache.cassandra.service.ClientState; +import org.apache.cassandra.service.QueryState; +import org.apache.cassandra.service.StorageService; +import org.apache.cassandra.transport.Dispatcher; +import org.apache.cassandra.transport.ProtocolVersion; +import org.apache.cassandra.transport.messages.ResultMessage; +import org.apache.cassandra.utils.ByteBufferUtil; +import org.apache.cassandra.utils.FBUtilities; +import org.apache.cassandra.repair.autorepair.AutoRepairConfig.RepairType; +import org.apache.cassandra.utils.NoSpamLogger; +import org.apache.cassandra.utils.concurrent.Refs; + +import static org.apache.cassandra.repair.autorepair.AutoRepairUtils.RepairTurn.MY_TURN; +import static org.apache.cassandra.repair.autorepair.AutoRepairUtils.RepairTurn.MY_TURN_DUE_TO_PRIORITY; +import static org.apache.cassandra.repair.autorepair.AutoRepairUtils.RepairTurn.NOT_MY_TURN; +import static org.apache.cassandra.repair.autorepair.AutoRepairUtils.RepairTurn.MY_TURN_FORCE_REPAIR; +import static org.apache.cassandra.utils.Clock.Global.currentTimeMillis; + +/** + * This class serves as a utility class for AutoRepair. It contains various helper APIs + * to store/retrieve repair status, decide whose turn is next, etc. + */ +public class AutoRepairUtils +{ + private static final Logger logger = LoggerFactory.getLogger(AutoRepairUtils.class); + static final String COL_REPAIR_TYPE = "repair_type"; + static final String COL_HOST_ID = "host_id"; + static final String COL_REPAIR_START_TS = "repair_start_ts"; + static final String COL_REPAIR_FINISH_TS = "repair_finish_ts"; + static final String COL_REPAIR_PRIORITY = "repair_priority"; + static final String COL_DELETE_HOSTS = "delete_hosts"; // this set stores the host ids which think the row should be deleted + static final String COL_REPAIR_TURN = "repair_turn"; // this record the last repair turn. Normal turn or turn due to priority + static final String COL_DELETE_HOSTS_UPDATE_TIME = "delete_hosts_update_time"; // the time when delete hosts are upated + static final String COL_FORCE_REPAIR = "force_repair"; // if set to true, the node will do non-primary range rapair + + final static String SELECT_REPAIR_HISTORY = String.format( + "SELECT * FROM %s.%s WHERE %s = ?", SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, + SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_REPAIR_TYPE); + final static String SELECT_REPAIR_PRIORITY = String.format( + "SELECT * FROM %s.%s WHERE %s = ?", SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, + SystemDistributedKeyspace.AUTO_REPAIR_PRIORITY, COL_REPAIR_TYPE); + final static String DEL_REPAIR_PRIORITY = String.format( + "DELETE %s[?] FROM %s.%s WHERE %s = ?", COL_REPAIR_PRIORITY, SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, + SystemDistributedKeyspace.AUTO_REPAIR_PRIORITY, COL_REPAIR_TYPE); + final static String ADD_PRIORITY_HOST = String.format( + "UPDATE %s.%s SET %s = %s + ? WHERE %s = ?", SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, + SystemDistributedKeyspace.AUTO_REPAIR_PRIORITY, COL_REPAIR_PRIORITY, COL_REPAIR_PRIORITY, COL_REPAIR_TYPE); + + final static String INSERT_NEW_REPAIR_HISTORY = String.format( + "INSERT INTO %s.%s (%s, %s, %s, %s, %s, %s) values (?, ? ,?, ?, {}, ?) IF NOT EXISTS", + SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_REPAIR_TYPE, + COL_HOST_ID, COL_REPAIR_START_TS, COL_REPAIR_FINISH_TS, COL_DELETE_HOSTS, COL_DELETE_HOSTS_UPDATE_TIME); + + final static String ADD_HOST_ID_TO_DELETE_HOSTS = String.format( + "UPDATE %s.%s SET %s = %s + ?, %s = ? WHERE %s = ? AND %s = ? IF EXISTS" + , SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_DELETE_HOSTS, + COL_DELETE_HOSTS, COL_DELETE_HOSTS_UPDATE_TIME, COL_REPAIR_TYPE, COL_HOST_ID); + + final static String DEL_AUTO_REPAIR_HISTORY = String.format( + "DELETE FROM %s.%s WHERE %s = ? AND %s = ?" + , SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_REPAIR_TYPE, + COL_HOST_ID); + + final static String RECORD_START_REPAIR_HISTORY = String.format( + "UPDATE %s.%s SET %s= ?, repair_turn = ? WHERE %s = ? AND %s = ?" + , SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_REPAIR_START_TS, + COL_REPAIR_TYPE, COL_HOST_ID); + + final static String RECORD_FINISH_REPAIR_HISTORY = String.format( + "UPDATE %s.%s SET %s= ?, %s=false WHERE %s = ? AND %s = ?" + , SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_REPAIR_FINISH_TS, + COL_FORCE_REPAIR, COL_REPAIR_TYPE, COL_HOST_ID); + + final static String CLEAR_DELETE_HOSTS = String.format( + "UPDATE %s.%s SET %s= {} WHERE %s = ? AND %s = ?" + , SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_DELETE_HOSTS, + COL_REPAIR_TYPE, COL_HOST_ID); + + final static String SET_FORCE_REPAIR = String.format( + "UPDATE %s.%s SET %s=true WHERE %s = ? AND %s = ?" + , SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_FORCE_REPAIR, + COL_REPAIR_TYPE, COL_HOST_ID); + + final static String SELECT_LAST_REPAIR_TIME_FOR_NODE = String.format( + "SELECT %s FROM %s.%s WHERE %s = ? AND %s = ?", COL_REPAIR_FINISH_TS, SchemaConstants.DISTRIBUTED_KEYSPACE_NAME, + SystemDistributedKeyspace.AUTO_REPAIR_HISTORY, COL_REPAIR_TYPE, COL_HOST_ID); + + static ModificationStatement delStatementRepairHistory; + static SelectStatement selectStatementRepairHistory; + static ModificationStatement delStatementPriorityStatus; + static SelectStatement selectStatementRepairPriority; + static SelectStatement selectLastRepairTimeForNode; + static ModificationStatement addPriorityHost; + static ModificationStatement insertNewRepairHistoryStatement; + static ModificationStatement recordStartRepairHistoryStatement; + static ModificationStatement recordFinishRepairHistoryStatement; + static ModificationStatement addHostIDToDeleteHostsStatement; + static ModificationStatement clearDeleteHostsStatement; + static ModificationStatement setForceRepairStatement; + static ConsistencyLevel internalQueryCL; + + public enum RepairTurn + { + MY_TURN, + NOT_MY_TURN, + MY_TURN_DUE_TO_PRIORITY, + MY_TURN_FORCE_REPAIR + } + + public static void setup() + { + selectStatementRepairHistory = (SelectStatement) QueryProcessor.getStatement(SELECT_REPAIR_HISTORY, ClientState + .forInternalCalls()); + selectStatementRepairPriority = (SelectStatement) QueryProcessor.getStatement(SELECT_REPAIR_PRIORITY, ClientState + .forInternalCalls()); + selectLastRepairTimeForNode = (SelectStatement) QueryProcessor.getStatement(SELECT_LAST_REPAIR_TIME_FOR_NODE, ClientState + .forInternalCalls()); + delStatementPriorityStatus = (ModificationStatement) QueryProcessor.getStatement(DEL_REPAIR_PRIORITY, ClientState + .forInternalCalls()); + addPriorityHost = (ModificationStatement) QueryProcessor.getStatement(ADD_PRIORITY_HOST, ClientState + .forInternalCalls()); + insertNewRepairHistoryStatement = (ModificationStatement) QueryProcessor.getStatement(INSERT_NEW_REPAIR_HISTORY, ClientState + .forInternalCalls()); + recordStartRepairHistoryStatement = (ModificationStatement) QueryProcessor.getStatement(RECORD_START_REPAIR_HISTORY, ClientState + .forInternalCalls()); + recordFinishRepairHistoryStatement = (ModificationStatement) QueryProcessor.getStatement(RECORD_FINISH_REPAIR_HISTORY, ClientState + .forInternalCalls()); + addHostIDToDeleteHostsStatement = (ModificationStatement) QueryProcessor.getStatement(ADD_HOST_ID_TO_DELETE_HOSTS, ClientState + .forInternalCalls()); + setForceRepairStatement = (ModificationStatement) QueryProcessor.getStatement(SET_FORCE_REPAIR, ClientState + .forInternalCalls()); + clearDeleteHostsStatement = (ModificationStatement) QueryProcessor.getStatement(CLEAR_DELETE_HOSTS, ClientState + .forInternalCalls()); + delStatementRepairHistory = (ModificationStatement) QueryProcessor.getStatement(DEL_AUTO_REPAIR_HISTORY, ClientState + .forInternalCalls()); + Keyspace autoRepairKS = Schema.instance.getKeyspaceInstance(SchemaConstants.DISTRIBUTED_KEYSPACE_NAME); + internalQueryCL = autoRepairKS.getReplicationStrategy().getClass() == NetworkTopologyStrategy.class ? + ConsistencyLevel.LOCAL_QUORUM : ConsistencyLevel.ONE; + } + + public static class AutoRepairHistory + { + UUID hostId; + String repairTurn; + long lastRepairStartTime; + long lastRepairFinishTime; + Set<UUID> deleteHosts; + long deleteHostsUpdateTime; + boolean forceRepair; + + public AutoRepairHistory(UUID hostId, String repairTurn, long lastRepairStartTime, long lastRepairFinishTime, + Set<UUID> deleteHosts, long deleteHostsUpateTime, boolean forceRepair) + { + this.hostId = hostId; + this.repairTurn = repairTurn; + this.lastRepairStartTime = lastRepairStartTime; + this.lastRepairFinishTime = lastRepairFinishTime; + this.deleteHosts = deleteHosts; + if (this.deleteHosts == null) + { + this.deleteHosts = new HashSet<>(); + } + this.deleteHostsUpdateTime = deleteHostsUpateTime; + this.forceRepair = forceRepair; + } + + public String toString() + { + return MoreObjects.toStringHelper(this). + add("hostId", hostId). + add("repairTurn", repairTurn). + add("lastRepairStartTime", lastRepairStartTime). + add("lastRepairFinishTime", lastRepairFinishTime). + add("deleteHosts", deleteHosts). + toString(); + } + + public boolean isRepairRunning() + { + // if a repair history record has start time laster than finish time, it means the repair is running + return lastRepairStartTime > lastRepairFinishTime; + } + + public long getLastRepairFinishTime() + { + return lastRepairFinishTime; + } + } + + public static class CurrentRepairStatus + { + public Set<UUID> hostIdsWithOnGoingRepair; // hosts that is running repair + public Set<UUID> hostIdsWithOnGoingForceRepair; // hosts that is running repair because of force repair + Set<UUID> priority; + public AutoRepairHistory myRepairHistory; + List<AutoRepairHistory> historiesWithoutOnGoingRepair; // hosts that is NOT running repair + + public CurrentRepairStatus(List<AutoRepairHistory> repairHistories, Set<UUID> priority, UUID myId) + { + hostIdsWithOnGoingRepair = new HashSet<>(); + hostIdsWithOnGoingForceRepair = new HashSet<>(); + historiesWithoutOnGoingRepair = new ArrayList<>(); + + for (AutoRepairHistory history : repairHistories) + { + if (history.isRepairRunning()) + { + if (history.forceRepair) + { + hostIdsWithOnGoingForceRepair.add(history.hostId); + } + else + { + hostIdsWithOnGoingRepair.add(history.hostId); + } + } + else + { + historiesWithoutOnGoingRepair.add(history); + } + if (history.hostId.equals(myId)) + { + myRepairHistory = history; + } + } + this.priority = priority; + } + + public Set<UUID> getAllHostsWithOngoingRepair() + { + return Sets.union(hostIdsWithOnGoingRepair, hostIdsWithOnGoingForceRepair); + } + + public String toString() + { + return MoreObjects.toStringHelper(this). + add("hostIdsWithOnGoingRepair", hostIdsWithOnGoingRepair). + add("hostIdsWithOnGoingForceRepair", hostIdsWithOnGoingForceRepair). + add("historiesWithoutOnGoingRepair", historiesWithoutOnGoingRepair). + add("priority", priority). + add("myRepairHistory", myRepairHistory). + toString(); + } + } + + @VisibleForTesting + public static List<AutoRepairHistory> getAutoRepairHistory(RepairType repairType) + { + UntypedResultSet repairHistoryResult; + + ResultMessage.Rows repairStatusRows = selectStatementRepairHistory.execute(QueryState.forInternalCalls(), + QueryOptions.forInternalCalls(internalQueryCL, Lists.newArrayList(ByteBufferUtil.bytes(repairType.toString()))), Dispatcher.RequestTime.forImmediateExecution()); + repairHistoryResult = UntypedResultSet.create(repairStatusRows.result); + + List<AutoRepairHistory> repairHistories = new ArrayList<>(); + if (!repairHistoryResult.isEmpty()) + { + for (UntypedResultSet.Row row : repairHistoryResult) + { + UUID hostId = row.getUUID(COL_HOST_ID); + String repairTurn = null; + if (row.has(COL_REPAIR_TURN)) + repairTurn = row.getString(COL_REPAIR_TURN); + long lastRepairStartTime = row.getLong(COL_REPAIR_START_TS, 0); + long lastRepairFinishTime = row.getLong(COL_REPAIR_FINISH_TS, 0); + Set<UUID> deleteHosts = row.getSet(COL_DELETE_HOSTS, UUIDType.instance); + long deleteHostsUpdateTime = row.getLong(COL_DELETE_HOSTS_UPDATE_TIME, 0); + boolean forceRepair = row.has(COL_FORCE_REPAIR) && row.getBoolean(COL_FORCE_REPAIR); + repairHistories.add(new AutoRepairHistory(hostId, repairTurn, lastRepairStartTime, lastRepairFinishTime, + deleteHosts, deleteHostsUpdateTime, forceRepair)); + } + return repairHistories; + } + logger.info("No repair history found"); + return null; + } + + // A host may add itself in delete hosts for some other hosts due to restart or some temp gossip issue. If a node's record + // delete_hosts is not growing for more than 2 hours, we consider it as a normal node so we clear the delete_hosts for that node + public static void clearDeleteHosts(RepairType repairType, UUID hostId) + { + clearDeleteHostsStatement.execute(QueryState.forInternalCalls(), + QueryOptions.forInternalCalls(internalQueryCL, + Lists.newArrayList(ByteBufferUtil.bytes(repairType.toString()), + ByteBufferUtil.bytes(hostId))), Dispatcher.RequestTime.forImmediateExecution()); + } + + public static void setForceRepairNewNode(RepairType repairType) + { + // this function will be called when a node bootstrap finished + UUID hostId = Gossiper.instance.getHostId(FBUtilities.getBroadcastAddressAndPort()); + // insert the data first + insertNewRepairHistory(repairType, currentTimeMillis(), currentTimeMillis()); + setForceRepair(repairType, hostId); + } + + public static void setForceRepair(RepairType repairType, Set<InetAddressAndPort> hosts) + { + // this function is used by nodetool + for (InetAddressAndPort host : hosts) + { + UUID hostId = Gossiper.instance.getHostId(host); + setForceRepair(repairType, hostId); + } + } + + public static void setForceRepair(RepairType repairType, UUID hostId) + { + setForceRepairStatement.execute(QueryState.forInternalCalls(), + QueryOptions.forInternalCalls(internalQueryCL, + Lists.newArrayList(ByteBufferUtil.bytes(repairType.toString()), + ByteBufferUtil.bytes(hostId))), + Dispatcher.RequestTime.forImmediateExecution()); + + logger.info("Set force repair repair type: {}, node: {}", repairType, hostId); + } + + public static long getLastRepairTimeForNode(RepairType repairType, UUID hostId) + { + ResultMessage.Rows rows = selectLastRepairTimeForNode.execute(QueryState.forInternalCalls(), + QueryOptions.forInternalCalls(internalQueryCL, + Lists.newArrayList( + ByteBufferUtil.bytes(repairType.toString()), + ByteBufferUtil.bytes(hostId))), + Dispatcher.RequestTime.forImmediateExecution()); + UntypedResultSet repairTime = UntypedResultSet.create(rows.result); + if (repairTime.isEmpty()) + { + return 0; + } + return repairTime.one().getLong(COL_REPAIR_FINISH_TS); + } + + @VisibleForTesting + public static CurrentRepairStatus getCurrentRepairStatus(RepairType repairType, List<AutoRepairHistory> autoRepairHistories, UUID myId) + { + if (autoRepairHistories != null) + { + return new CurrentRepairStatus(autoRepairHistories, getPriorityHostIds(repairType), myId); + } + return null; + } + + /** + * Checks whether the cluster has multiple major versions + * @return + * true if more than one major versions are detected + * false if only one major version is detected + * + */ + public static boolean hasMultipleLiveMajorVersions() + { + Set<InetAddressAndPort> liveEndpoints = Gossiper.instance.getLiveMembers(); + Set<Integer> majorVersions = new HashSet<>(); + for (InetAddressAndPort endpoint : liveEndpoints) + { + CassandraVersion releaseVersion = Gossiper.instance.getReleaseVersion(endpoint); + if (releaseVersion != null) + { + majorVersions.add(releaseVersion.major); + } + } + return majorVersions.size() > 1; + } + + /** + * Last version that does not support auto-repair. + * All nodes in the cluster must be running a version above this to enable auto-repair. + * Versions at or below this version (5.0.6) do not support auto-repair. + */ + @VisibleForTesting + static final CassandraVersion LAST_UNSUPPORTED_VERSION_FOR_AUTO_REPAIR = new CassandraVersion("5.0.6"); + + /** + * Checks whether any node in the cluster is running an unsupported version for auto-repair. + * + * @return true if any live node has a version at or below 5.0.6 (unsupported) or has an unknown version, + * false if all nodes are running versions above 5.0.6 (supported) + */ + public static boolean hasNodesBelowMinimumVersion() + { + Set<InetAddressAndPort> liveEndpoints = Gossiper.instance.getLiveMembers(); + for (InetAddressAndPort endpoint : liveEndpoints) + { + CassandraVersion releaseVersion = Gossiper.instance.getReleaseVersion(endpoint); + if (releaseVersion == null) + { + logger.warn("Cannot determine version for endpoint {}, blocking auto-repair", endpoint); + return true; + } + if (releaseVersion.compareTo(LAST_UNSUPPORTED_VERSION_FOR_AUTO_REPAIR) <= 0) + { + logger.info("Endpoint {} is running version {} which does not support auto-repair " + + "(auto-repair requires version above {})", + endpoint, releaseVersion, LAST_UNSUPPORTED_VERSION_FOR_AUTO_REPAIR); + return true; + } + } + return false; + } + + @VisibleForTesting + protected static TreeSet<UUID> getHostIdsInCurrentRing(RepairType repairType, Set<InetAddressAndPort> allNodesInRing) + { + TreeSet<UUID> hostIdsInCurrentRing = new TreeSet<>(); + for (InetAddressAndPort node : allNodesInRing) + { + String nodeDC = DatabaseDescriptor.getEndpointSnitch().getDatacenter(node); + if (AutoRepairService.instance.getAutoRepairConfig().getIgnoreDCs(repairType).contains(nodeDC)) + { + logger.info("Ignore node {} because its datacenter is {}", node, nodeDC); + continue; + } + /* + * Check if endpoint state exists in gossip or not. If it + * does not then this maybe a ghost node so ignore it + */ + if (Gossiper.instance.isAlive(node)) + { + UUID hostId = Gossiper.instance.getHostId(node); + hostIdsInCurrentRing.add(hostId); + } + else + { + logger.warn("Node is not present in Gossip cache node {}, node data center {}", node, nodeDC); + } + } + return hostIdsInCurrentRing; + } + + public static TreeSet<UUID> getHostIdsInCurrentRing(RepairType repairType) + { + Set<InetAddressAndPort> allNodesInRing = StorageService.instance.getTokenMetadata().getAllEndpoints(); + return getHostIdsInCurrentRing(repairType, allNodesInRing); + } + + // This function will return the host ID for the node which has not been repaired for longest time + public static AutoRepairHistory getHostWithLongestUnrepairTime(RepairType repairType) + { + List<AutoRepairHistory> autoRepairHistories = getAutoRepairHistory(repairType); + return getHostWithLongestUnrepairTime(autoRepairHistories); + } + + /** + * Convenience method to resolve the broadcast address of a host id + * + * @return broadcast address if it exists, otherwise null. + */ + @Nullable + private static InetAddressAndPort getBroadcastAddress(UUID hostId) + { + return StorageService.instance.getEndpointForHostId(hostId); + } + + /** + * @return Map of broadcast address to host id, if a broadcast address cannot be found for a host, it is + * not included in the map. + */ + private static Map<InetAddressAndPort, UUID> getBroadcastAddressToHostIdMap(Set<UUID> hosts) + { + // Get a mapping of endpoint : host id + Map<InetAddressAndPort, UUID> broadcastAddressMap = new HashMap<>(hosts.size()); + for (UUID hostId : hosts) + { + InetAddressAndPort broadcastAddress = getBroadcastAddress(hostId); + if (broadcastAddress == null) + { + logger.warn("Could not resolve broadcast address from host id {} in ClusterMetadata can't accurately " + + "determine if this node is a replica of the local node.", hostId); + } + else + { + broadcastAddressMap.put(broadcastAddress, hostId); + } + } + return broadcastAddressMap; + } + + /** + * @return Mapping of unique replication strategy to keyspaces using that strategy that we care about repairing. + */ + private static Map<AbstractReplicationStrategy, List<String>> getReplicationStrategies() + { + // Collect all unique replication strategies among all keyspaces. + Map<AbstractReplicationStrategy, List<String>> replicationStrategies = new HashMap<>(); + for (Keyspace keyspace : Keyspace.all()) + { + if (AutoRepairUtils.shouldConsiderKeyspace(keyspace)) + { + replicationStrategies.computeIfAbsent(keyspace.getReplicationStrategy(), k -> new ArrayList<>()) + .add(keyspace.getName()); + } + } + return replicationStrategies; + } + + /** + * Collects all hosts being repaired among all active repair schedules and their schedule if + * {@link AutoRepairConfig#getAllowParallelReplicaRepairAcrossSchedules(RepairType)} is true for this repairType. + * Accepts the currently evaluated repairType's schedule as an optimization to avoid grabbing its repair status an + * additional time. + * + * @param myRepairType The repair type schedule being evaluated. + * @param myRepairStatus The repair status for that repair type. + * @return All hosts among active schedules currently being repaired. + */ + private static Map<UUID, RepairType> getHostsBeingRepaired(RepairType myRepairType, CurrentRepairStatus myRepairStatus) + { + AutoRepairConfig config = AutoRepairService.instance.getAutoRepairConfig(); + + Map<UUID, RepairType> hostsBeingRepaired = myRepairStatus.getAllHostsWithOngoingRepair().stream() + .collect(Collectors.toMap((h) -> h, (v) -> myRepairType)); + + // If we don't allow repairing across schedules, iterate over other enabled schedules and include hosts + // actively being repaired. + if (!config.getAllowParallelReplicaRepairAcrossSchedules(myRepairType)) + { + for (RepairType repairType : RepairType.values()) + { + if (myRepairType == repairType) + continue; + + if (config.isAutoRepairEnabled(repairType)) + { + CurrentRepairStatus repairStatus = getCurrentRepairStatus(repairType, getAutoRepairHistory(repairType), null); + if (repairStatus != null) + { + for (UUID hostId : repairStatus.getAllHostsWithOngoingRepair()) + { + hostsBeingRepaired.putIfAbsent(hostId, repairType); + } + } + } + } + } + return hostsBeingRepaired; + } + + /** + * Identifies the most eligible host to repair for nodes preceding or equal to this nodes' lastRepairFinishTime. + * The criteria for this is to find the node with the oldest last repair finish time of which none of its replicas + * are currently under repair. + * + * @return The most eligible host to repair or null if no candidates before and including this nodes' current repair status. + */ + @VisibleForTesting + public static AutoRepairHistory getMostEligibleHostToRepair(RepairType repairType, CurrentRepairStatus currentRepairStatus, UUID myId) + { + // 0. If this repairType allows parallel replica repair, short circuit and return the host with the longest unrepair time + AutoRepairConfig config = AutoRepairService.instance.getAutoRepairConfig(); + if (config.getAllowParallelReplicaRepair(repairType)) + { + return getHostWithLongestUnrepairTime(currentRepairStatus.historiesWithoutOnGoingRepair); + } + + // 1. Sort repair histories from oldest completed to newest + Stream<AutoRepairHistory> finishedRepairHistories = currentRepairStatus.historiesWithoutOnGoingRepair + .stream() + .sorted(Comparator.comparingLong(h -> h.lastRepairFinishTime)); + + // 2. Optimization: Truncate repair histories after myId so we don't evaluate anything more recent as if we + // aren't interested in anything that isn't this node. + final AtomicBoolean myHistoryFound = new AtomicBoolean(false); + finishedRepairHistories = finishedRepairHistories + .filter(history -> { + if (myHistoryFound.get()) return false; // Stop processing after finding myId + myHistoryFound.set(history.hostId.equals(myId)); + return true; + }); + + // If there are any hosts with ongoing repair, filter the repair histories to not include nodes whose replicas + // are ongoing repair. + Map<UUID, RepairType> hostsBeingRepairedToRepairType = getHostsBeingRepaired(repairType, currentRepairStatus); + + // 3. If I am already actively being repaired in another schedule, defer submitting repairs; if already + // repairing for this type, return node so it can take its turn. + RepairType alreadyRepairingType = hostsBeingRepairedToRepairType.get(myId); + if (alreadyRepairingType != null) + { + if (repairType != alreadyRepairingType) + { + logger.info("Deferring repair because I am already actively repairing in schedule {}", hostsBeingRepairedToRepairType.get(myId)); + AutoRepairMetricsManager.getMetrics(repairType).repairDelayedBySchedule.inc(); + return null; + } + else if (currentRepairStatus.myRepairHistory != null) + { + // if the repair type matches this repair, assume the node was restarted while repairing, return node + // so it can take its turn. + logAlreadyMyTurn(); + return currentRepairStatus.myRepairHistory; + } + } + + if (!hostsBeingRepairedToRepairType.isEmpty()) + { + // 4. Extract InetAddresses for each UUID as replicas are identified by their address. + Map<InetAddressAndPort, UUID> hostsBeingRepaired = getBroadcastAddressToHostIdMap(hostsBeingRepairedToRepairType.keySet()); + + // 5. Collect unique replication strategies and group them up with their keyspaces. + Map<AbstractReplicationStrategy, List<String>> replicationStrategies = getReplicationStrategies(); + + // 6. Filter out repair histories who have a replica being repaired, note that this is lazy, given the stream + // is completed using findFirst, it should stop as soon as the matching criteria is met. + finishedRepairHistories = finishedRepairHistories.filter((history) -> !hasReplicaWithOngoingRepair(history, + myId, + repairType, + hostsBeingRepaired, + hostsBeingRepairedToRepairType, + replicationStrategies)); + } + + // 7. Select the first (oldest lastRepairFinishTime) repair history without replicas being repaired + return finishedRepairHistories.findFirst().orElse(null); + } + + + /** + * @param eligibleHistory History of node to check + * @param myId Host id of this node, if the repair history is for this node, additional logging will take place. + * @param myRepairType repair type being evaluated + * @param hostsBeingRepaired Hosts being repaired. + * @param hostIdToRepairType mapping of hosts being repaired to the repair type its being repaired for. + * @param replicationStrategies Mapping of unique replication strategies to keyspaces having that strategy. + * @return Whether the host for the given eligibleRepairHistory has any replicas in hostsBeingRepaired. + */ + private static boolean hasReplicaWithOngoingRepair(AutoRepairHistory eligibleHistory, + UUID myId, + RepairType myRepairType, + Map<InetAddressAndPort, UUID> hostsBeingRepaired, + Map<UUID, RepairType> hostIdToRepairType, + Map<AbstractReplicationStrategy, List<String>> replicationStrategies) + { + // If no broadcast address found for this host id in cluster metadata, just skip it, a node should always + // see itself in cluster metadata. + InetAddressAndPort eligibleBroadcastAddress = getBroadcastAddress(eligibleHistory.hostId); + if (eligibleBroadcastAddress == null) + { + return true; + } + + // For each replication strategy, determine if host being repaired is a replica of the local node. + for (Map.Entry<AbstractReplicationStrategy, List<String>> entry : replicationStrategies.entrySet()) + { + AbstractReplicationStrategy replicationStrategy = entry.getKey(); + EndpointsByRange endpointsByRange = replicationStrategy.getRangeAddresses(StorageService.instance.getTokenMetadata()); + + // get ranges of the eligible address for the given replication strategy. + RangesAtEndpoint rangesAtEndpoint = StorageService.instance.getReplicas(replicationStrategy, eligibleBroadcastAddress); + for (Replica replica : rangesAtEndpoint) + { + // get the endpoints involved in this range. + EndpointsForRange endpointsForRange = endpointsByRange.get(replica.range()); + // For each host in this range... + for (InetAddressAndPort inetAddressAndPort : endpointsForRange.endpoints()) + { + // If the address of the node in the range belongs to a host being repaired, return true. + UUID hostId = hostsBeingRepaired.get(inetAddressAndPort); + if (hostId != null) + { + // log if the repair history matches the current running node. + InetAddressAndPort myBroadcastAddress = getBroadcastAddress(myId); + if (myBroadcastAddress != null && myBroadcastAddress.equals(eligibleBroadcastAddress)) + { + logger.info("Deferring repair because replica {} ({}) with shared ranges for " + + "{} keyspace(s) (e.g. {}) is currently taking its turn for schedule {}", + hostId, inetAddressAndPort, entry.getValue().size(), entry.getValue().get(0), + hostIdToRepairType.get(hostId)); + AutoRepairMetricsManager.getMetrics(myRepairType).repairDelayedByReplica.inc(); + } + else if (logger.isDebugEnabled()) + { + logger.debug("Not considering node {} ({}) for repair as it has replica {} ({}) with " + + "shared ranges for {} keyspace(s) (e.g. {}) which is currently taking its " + + "turn for schedule {}", + eligibleHistory.hostId, eligibleBroadcastAddress, + hostId, inetAddressAndPort, entry.getValue().size(), entry.getValue().get(0), + hostIdToRepairType.get(hostId)); + } + return true; + } + } + } + } + + // No replicas found of eligible host. + return false; + } + + private static AutoRepairHistory getHostWithLongestUnrepairTime(List<AutoRepairHistory> autoRepairHistories) + { + if (autoRepairHistories == null) + { + return null; + } + AutoRepairHistory rst = null; + long oldestTimestamp = Long.MAX_VALUE; + for (AutoRepairHistory autoRepairHistory : autoRepairHistories) + { + if (autoRepairHistory.lastRepairFinishTime < oldestTimestamp) + { + rst = autoRepairHistory; + oldestTimestamp = autoRepairHistory.lastRepairFinishTime; + } + } + return rst; + } + + public static int getMaxNumberOfNodeRunAutoRepair(RepairType repairType, int groupSize) + { + AutoRepairConfig config = AutoRepairService.instance.getAutoRepairConfig(); + if (groupSize == 0) + { + return Math.max(config.getParallelRepairCount(repairType), 1); + } + // we will use the max number from config between auto_repair_parallel_repair_count_in_group and auto_repair_parallel_repair_percentage_in_group + int value = Math.max(groupSize * config.getParallelRepairPercentage(repairType) / 100, + config.getParallelRepairCount(repairType)); + // make sure at least one node getting repaired + return Math.max(1, value); + } + + private static void logAlreadyMyTurn() + { + logger.warn("This node already was considered to having an ongoing repair for this repair type, must have " + + "been restarted, taking my turn back"); + } + + @VisibleForTesting + public static RepairTurn myTurnToRunRepair(RepairType repairType, UUID myId) + { + try + { + Set<InetAddressAndPort> allNodesInRing = StorageService.instance.getTokenMetadata().getAllEndpoints(); + logger.info("Total nodes in ring {}", allNodesInRing.size()); + TreeSet<UUID> hostIdsInCurrentRing = getHostIdsInCurrentRing(repairType, allNodesInRing); + logger.info("Total nodes qualified for repair {}", hostIdsInCurrentRing.size()); + + List<AutoRepairHistory> autoRepairHistories = getAutoRepairHistory(repairType); + Set<UUID> autoRepairHistoryIds = new HashSet<>(); + + // 1. Remove any node that is not part of group based on gossip info + if (autoRepairHistories != null) + { + for (AutoRepairHistory nodeHistory : autoRepairHistories) + { + autoRepairHistoryIds.add(nodeHistory.hostId); + // clear delete_hosts if the node's delete hosts is not growing for more than two hours + AutoRepairConfig config = AutoRepairService.instance.getAutoRepairConfig(); + if (!nodeHistory.deleteHosts.isEmpty() + && config.getAutoRepairHistoryClearDeleteHostsBufferInterval().toSeconds() < TimeUnit.MILLISECONDS.toSeconds( + currentTimeMillis() - nodeHistory.deleteHostsUpdateTime + )) + { + clearDeleteHosts(repairType, nodeHistory.hostId); + logger.info("Delete hosts for {} for repair type {} has not been updated for more than {} seconds. Delete hosts has been cleared. Delete hosts before clear {}" + , nodeHistory.hostId, repairType, config.getAutoRepairHistoryClearDeleteHostsBufferInterval(), nodeHistory.deleteHosts); + } + else if (!hostIdsInCurrentRing.contains(nodeHistory.hostId)) + { + if (nodeHistory.deleteHosts.size() > Math.max(2, hostIdsInCurrentRing.size() * 0.5)) + { + // More than half of the groups thinks the record should be deleted + logger.info("{} think {} is orphan node, will delete auto repair history for repair type {}.", nodeHistory.deleteHosts, nodeHistory.hostId, repairType); + deleteAutoRepairHistory(repairType, nodeHistory.hostId); + } + else + { + // I think this host should be deleted + logger.info("I({}) think {} is not part of ring, vote to delete it for repair type {}.", myId, nodeHistory.hostId, repairType); + addHostIdToDeleteHosts(repairType, myId, nodeHistory.hostId); + } + } + } + } + + // 2. Add node to auto repair history table if a node is in gossip info + for (UUID hostId : hostIdsInCurrentRing) + { + if (!autoRepairHistoryIds.contains(hostId)) + { + logger.info("{} for repair type {} doesn't exist in the auto repair history table, insert a new record.", repairType, hostId); + insertNewRepairHistory(repairType, hostId, currentTimeMillis(), currentTimeMillis()); + } + } + + // get updated current repair status + CurrentRepairStatus currentRepairStatus = getCurrentRepairStatus(repairType, getAutoRepairHistory(repairType), myId); + if (currentRepairStatus != null) + { + if (logger.isDebugEnabled()) + { + logger.debug("Latest repair status {}", currentRepairStatus); + } + //check if I am forced to run repair + for (AutoRepairHistory history : currentRepairStatus.historiesWithoutOnGoingRepair) + { + if (history.forceRepair && history.hostId.equals(myId)) + { + return MY_TURN_FORCE_REPAIR; + } + } + } + + // check if node was already indicated as having an ongoing repair, this may happen when a node restarts + // before finishing repairing. + if (currentRepairStatus != null && currentRepairStatus.getAllHostsWithOngoingRepair().contains(myId)) + { + logAlreadyMyTurn(); + + // use the previously chosen turn. + if (currentRepairStatus.myRepairHistory != null && currentRepairStatus.myRepairHistory.repairTurn != null) + { + return RepairTurn.valueOf(currentRepairStatus.myRepairHistory.repairTurn); + } + else + { + return MY_TURN; + } + } + + int parallelRepairNumber = getMaxNumberOfNodeRunAutoRepair(repairType, + autoRepairHistories == null ? 0 : autoRepairHistories.size()); + logger.info("Will run repairs concurrently on {} node(s)", parallelRepairNumber); + if (currentRepairStatus == null || parallelRepairNumber > currentRepairStatus.hostIdsWithOnGoingRepair.size()) + { + // more repairs can be run, I might be the new one + if (autoRepairHistories != null) Review Comment: @jaydeepkumar1984 do you recall why do we re-fetch the repair history here? In what circumstances can `autoRepairHistories` can be null, is this just on an uninitialized system ? I am wondering if this could cause a race where multiple nodes will attempt to run repair more than the allowed parallelNumber. Should we not perform this re-fetch outside the `if parallelRepairNumber > currentRepairStatus.hostIdsWithOnGoingRepair.size()` ? -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
